DOCSLIB.ORG

Back to the Future Part 4: the Internet

Total Page：16

File Type：pdf, Size：1020Kb

Download full-text PDF Read full-text
Back to the Future Part 4: the Internet ∗ Back to the Future Part 4: The Internet Simon S. Lam Department of Computer Sciences The University of Texas at Austin [email protected] Slide 1 I was much younger then. The great thing about young people is that they have no fear. I don’t remember that I ever worried about losing money. I arranged to use the Thompson conference center on campus which charged us only a nominal fee. We had a great lineup of speakers. The pre-conference tutorial on network protocol design was by David Clark, a former SIGCOMM Award winner. The Keynote session had three speakers: Vint Cerf, another for- mer SIGCOMM Award winner, John Shoch of Xerox who Back to the Future Part 4: was in charge of Ethernet development at that time, and The Internet Louis Pouzin, yet another former SIGCOMM Award win- ner. We had 220 attendees, an excellent attendance for a first-time conference. We ended up making quite a bit of Simon S. Lam money.Endofstory. Department of Computer Sciences The University of Texas at Austin Slide 2 Sigcomm 2004 Keynote 1 S. S. Lam IP won the networking race I thank the SIGCOMM Awards committee for this great Many competitors in the honor. I would like to share this honor with my former and FTP SMTP HTTP RTP DNS current students, and colleagues, who collaborated and co- past authored papers with me. Each one of them has contributed SNA, DECnet, XNS to make this award possible. To all my coauthors, I thank X.25, ATM, CLNP TCP UDP you. It is most gratifying for me to accept this award at a SIG- IP provides end-to-end IP COMM conference because my relationship with the confer- delivery of datagrams, ence goes back a long way. Over 21 years ago, I organized best-effort service only . the very first SIGCOMM conference held on the campus of IP can use any link-layer LINK1 LINK2 LINKn the University of Texas at Austin in March 1983. Let me technology that delivers tell you a little story about the first SIGCOMM. datagrams (packets) Back in 1982–1983, the SIGCOMM chair was David Wood. Sigcomm 2004 Keynote 2 The vice chair was Carl Sunshine. One day during the sum- S. S. Lam mer of 1982, Carl called me up on the phone. He said, “Si- mon, SIGCOMM is broke. We have a lot of red ink. How IP won the networking race for data communications. would you like to organize a SIGCOMM conference to make However, as recently as ten years ago, before widespread some money for us?” I said, “Okay, I will do it.” Then, we use of the World Wide Web, it was not clear that IP would talked a bit. Towards the end of our conversation, Carl said, be the winner. The original ARPANET in the 1970s and “By the way, Simon, as I said, SIGCOMM is broke. So if later the Internet had many competitors in the past. you lose any money on the conference, you will have to take In industry, the competitors included SNA, IBM’s Sys- care of the loss yourself.” tems Network Architecture, DECnet with the Digital Net- work Architecture, and XNS which stands for Xerox Net- ∗This is an edited transcript of the author’s SIGCOMM 2004 keynote speech given on August 31, 2004. This work was work systems. SNA was developed around 1975. At that sponsored in part by Nation Science Foundation grants ANI- time, IBM was the 500-pound gorilla in the computer indus- 0319168 and CNS-0434515 and Texas Advanced Research try. IBM already had networking products used by numer- Program grant 003658-0439-2001. ous customers. Those early networks had a tree topology with a mainframe computer connected to data concentra- packets may be discarded because of buffer overflow. As a tors and terminals. But SNA was designed to have a new result, a network of queues is prone to congestion collapse, as architecture with a mesh topology similar to ARPANET’s illustrated in this figure, where system throughput decreases architecture. Back in 1975, very few people would have pre- rapidly as offered load becomes large. This is a weakness of dicted that the small, experimental ARPANET, rather than IP that we should keep in mind. SNA, would emerge 25 years later as winner of the network- ing race. In fact, when SNA was first developed, SNA was Slide 4 the acronym for Single Network Architecture. IBM had very ambitious goals. However, by the time SNA was announced to the public, the name was changed to Systems Network Architecture, possibly due to antitrust concerns. IP also had strong competitors in the standards world. Congestion collapse—ALOHA channel First, there was X.25 in the 1970s and 1980s. Then there was ATM in the 1990s. Also, there were ISO protocols such The ALOHA System (Abramson 1970) as CLNP (Connectionless Network Protocol). Poisson process assumption IP is a very simple protocol. It provides end-to-end deliv- ery of datagrams, also called packets. It provides no service pure ALOHA throughput guarantee to its users. In turn, IP expects no service guar- S = G e-2G (Abramson) antee from any link-layer technology it uses. Therefore, any slotted ALOHA throughput network can connect to IP. S = G e-G (Roberts) Ten years ago, when Internet applications were primarily email, ftp, and web, IP’s simplicity was its greatest strength in fighting off competitors. In the future, however, IP’s ARPANET Satellite System (1972) simplicity is possibly a liability because the requirements of Internet’s future applications will be more demanding, Sigcomm 2004 Keynote 4 particularly the requirements of interactive multimedia ap- S. S. Lam plications. Congestion collapse was a subject of great interest to me Slide 3 when I was a Ph.D. student at UCLA. Congestion collapse of the ALHOA channel was the inspiration of my disserta- tion research. In the next several slides, I will show you a little bit of my early work in the 1970s. My early work had IP‛s underlying model is a network a lot of influence on my thinking, and may explain why I of queues hold certain opinions later on in this talk. I went to UCLA as a graduate student in 1969 when the Revolutionary change Unreliable channels, limited buffer capacity first IMP of the ARPANET was being installed there. The from the circuit model next year, Norm Abramson presented his seminal paper on (Kleinrock 1961) Prone to congestion collapse the ALOHA System in the Fall Joint Computer Conference. Abramson made use of a Poisson process assumption and t Each packet is routed u p obtained the well-known formula for the throughput of pure independently using its h g destination IP address u ALOHA. o r No concept of a flow h The throughput formula for slotted ALOHA was an obser- T between source and m vation by Larry Roberts in 1972. By then, the ARPANET e destination, no flow t s state in routers y was already up and running. Roberts was looking for some- S thing else to do. Motivated by the ALOHA System, he Offered Load started the ARPANET Satellite System project, later re- named the Packet Satellite project. I was one of the first Sigcomm 2004 Keynote 3 S. S. Lam graduate students to work on the project. Others include Bob Metcalfe who was then at Xerox PARC and Dick Binder IP’s underlying model is a network of queues. Packets are who was working on the ALOHA System. transmitted from one node to another, leaving one queue and joining the next queue as they travel from source to Slide 5 destination. Each packet travels independently. There is no concept of a flow of packets belonging to the same session Congestion collapse was not a possibility discussed by between source and destination processes. ALOHA System researchers. In their early presentations, I give credit to Len Kleinrock for proposing the network of the ALOHA System was always said to work well. This queues model for data communications as an alternative to is because the ALOHA channel was vastly under-utilized. the circuit model in telephone networks. Some might argue In today’s terminology, it was over-provisioned. One of my that Kleinrock did not invent the name “packet switching.” opinions is that over-provisioning covers up problems. But he was definitely the first to propose the network of Also, with the Poisson process assumption, Abramson ab- queues model in his 1961 Ph.D. dissertation proposal. stracted away in his analysis the need for a backoff algo- In a real network, channels are unreliable. More impor- rithm. So the first thing Len Kleinrock and I did for slotted tantly, buffer capacity for queueing is limited. Therefore, ALOHA was to introduce a realistic backoff algorithm for tously after about 3,000 time slots. I believe that this was Congestion collapse—ALOHA channel the first experimental demonstration of congestion collapse (cont.) and ALOHA instability. The early ALOHA System was vastly under- Slide 7 utilized Backoff algorithm for slotted ALOHA Congestion collapse—ALOHA channel (Kleinrock-Lam 1973) (cont.) Retransmit a collided packet randomly into Adaptive backoff algorithm (Lam 1974): one of K future time slots Retransmit a packet with m previous collisions into K(m) slots, where K(m) is monotonically Poisson process assumption implies nondecreasing in m K ˳ Sigcomm 2004 Keynote 5 S. S. Lam collided packets. In our algorithm each collided packet is re- K(1) = 10 K transmitted randomly into one of future time slots, where K(m) = 150, m ˻ 2 K is a parameter.
Load more

Recommended publications
  • Packet-Voice Communication on an Ethernet Local Computer Network: an Experimental Study
    Packet-Voice Communication on an Ethernet Local Computer Network: an Experimental Study Timothy A. Gonsalves Packet-Voice Communication on an Ethernet Local Computer Network: an Experimental Study Timothy A. Gonsalves CSL·82·5 March 1982 [P82·00034] © Copyright Timothy A. Gonsalves 1982. All rights reserved. Abstract: Local computer networks have been used successfully for data applications such as file transfers for several years. Recently, there have been several proposals for using these networks for voice applications. This paper describes a simple voice protocol for use on a packet-switching local network. This protocol is used in an experimental study of the feasibility of using a 3 Mbps experimental Ethernet network for packet-voice communications. The study shows that with appropriately chosen parameters the experimental Ethernet is capable of supporting about 40 simultaneous 64-Kbps voice conversations with acceptable quality. This corresponds to a utilization of 95% of the network capacity. A version of this paper has been submitted for presentation at the 3rd International Conference on Distributed Computing Systems, Miami, Florida, October 1982. CR Categories: C.2.5, C.4 Key words and Phrases: CSMA/CD, Ethernet, local computer network, measurement, packet-voice, voice communication. Author's Affiliation: Computer Systems Laboratory, Department of Electrical Engineering, Stanford University, Stanford, California 94305. Xerox Corporation XEROX Palo Alto Research Centers 3333 Coyote Hill Road Palo Alto, California 94304 I. Introduction In the past few years. local computer networks for the interconnection of computers and shared resources within a small area such as a building or a campus have rapidly increased in popularity. These networks support applications such as file transfers and electronic mail between autonomous computers.
    [Show full text]
  • Reliable File Transfer Across a 10 Megabit Ethernet
    Rochester Institute of Technology RIT Scholar Works Theses 1984 Reliable file transfer across a 10 megabit ethernet Mark Van Dellon Follow this and additional works at: https://scholarworks.rit.edu/theses Recommended Citation Van Dellon, Mark, "Reliable file transfer across a 10 megabit ethernet" (1984). Thesis. Rochester Institute of Technology. Accessed from This Thesis is brought to you for free and open access by RIT Scholar Works. It has been accepted for inclusion in Theses by an authorized administrator of RIT Scholar Works. For more information, please contact [email protected]. Rochester Institute of Technology School of Computer Science and Technology Reliable File Transfer Ac ross A 10 Megabit Ethernet by Mark Van Dellon July 24, 1984 A thesis, submitted to The Faculty of the Computer Science and Technology, in partial fulfillment of the requirements for the degree of Master of Science in Computer Science Approved by: Roy Czernikowski Dr. Roy - Committee Head John Ellis 'Or. John Ellis Tong-han Chang Dr. Tong-han Chang Date fr-was? Copyright 1984. All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, electronic, mechanical, photocopying, recording, or otherwise, for commercial use or profit without the prior written permission of the author. Abstract: The Ethernet communications network is a broadcast, multi-access system for local computing networks. Such a network was used to connect six 68000 based Charles River Data Systems for the purpose of file transfer. Each system required hardware installation and connection to the Ethernet cable.
    [Show full text]
  • Network Traffic Modeling
    Chapter in The Handbook of Computer Networks, Hossein Bidgoli (ed.), Wiley, to appear 2007 Network Traffic Modeling Thomas M. Chen Southern Methodist University, Dallas, Texas OUTLINE: 1. Introduction 1.1. Packets, flows, and sessions 1.2. The modeling process 1.3. Uses of traffic models 2. Source Traffic Statistics 2.1. Simple statistics 2.2. Burstiness measures 2.3. Long range dependence and self similarity 2.4. Multiresolution timescale 2.5. Scaling 3. Continuous-Time Source Models 3.1. Traditional Poisson process 3.2. Simple on/off model 3.3. Markov modulated Poisson process (MMPP) 3.4. Stochastic fluid model 3.5. Fractional Brownian motion 4. Discrete-Time Source Models 4.1. Time series 4.2. Box-Jenkins methodology 5. Application-Specific Models 5.1. Web traffic 5.2. Peer-to-peer traffic 5.3. Video 6. Access Regulated Sources 6.1. Leaky bucket regulated sources 6.2. Bounding-interval-dependent (BIND) model 7. Congestion-Dependent Flows 7.1. TCP flows with congestion avoidance 7.2. TCP flows with active queue management 8. Conclusions 1 KEY WORDS: traffic model, burstiness, long range dependence, policing, self similarity, stochastic fluid, time series, Poisson process, Markov modulated process, transmission control protocol (TCP). ABSTRACT From the viewpoint of a service provider, demands on the network are not entirely predictable. Traffic modeling is the problem of representing our understanding of dynamic demands by stochastic processes. Accurate traffic models are necessary for service providers to properly maintain quality of service. Many traffic models have been developed based on traffic measurement data. This chapter gives an overview of a number of common continuous-time and discrete-time traffic models.
    [Show full text]
  • Ethernet LAN (Angolul Link Layer) NET
    Computer Networks Sándor Laki ELTE-Ericsson Communication Networks Laboratory ELTE FI – Department Of Information Systems [email protected] http://lakis.web.elte.hu Based on the slides of Laurent Vanbever. Further inspiration: Scott Shenker & Jennifer Rexford & Phillipa Gill Last week on Computer Networks Overview What is a network made of? Three main components End-points Switches Links Overview How to share network resources? Resource handling Two different approaches for sharing Reservation On-demand Reserve the needed Send data when needed bandwidth in advance Packet-level multiplexing Flow-level multiplexing Pros & Cons Pros Cons Predictable performance Low efficiency Bursty traffic Short flows Simple and fast switching Complexity of circuit establ./teard. once circuit established Increased delay New circuit is needed in case of failures Implementation Reservation On-demand Circuit-switching Packet-switching e.g. landline phone networks e.g. Internet Packets Overview How to organize the network? Tier-1 ISP Tier-1 ISP IXP Tier-2 ISP Tier-2 ISP Access ISP Access ISP This week How does communication happen? How do we characterize it? Briefly… The Internet should allow processes on different hosts to exchange data everything else is just commentary… Ok, but how to do that in a complex system like the Internet? University net Phone company CabelTV company Enterprise net To exchange data, Alice and Bob use a set of network protocols Alice Bob A protocol is like a conversational convention The protocol defines the order and rules the parties
    [Show full text]
  • Internet Research: Comments on Formulating the Problem
    Internet Research: Comments on Formulating the Problem Gathered by Sally Floyd, with contributions from Deborah Estrin, Greg Minshall, Vern Paxson, Lixia Zhang, and others. January 21, 1998 1 Introduction Development and deployment in the infrastructure is of necessity incremental. This note contains a discussion about formulating the research Explicit examined assumptions are better than implicit un- problem for Internet research. examined ones. The goal of this note is to further the discussion of implicit Changes in the Internet can be unanticipated and uncon- and explicit assumptions in network research. In particular, trolled. this note tries to articulate one such set of assumptions for In- The Internet architecture and scale make requirements for ternet research. Each of these assumptions is shared by some global consistency problematic. subset of the network research community, though perhaps Some research problems have their own natural time none of these assumptions are shared universally. The goal scales. of this paper is not to argue the validity of the assumptions, but to articulate them, to invite discussion of con¯icting or shared sets of assumptions, and to consider the implications 3 Discussion of the assumptions of these assumptions in formulating problems in Internet re- search. This process would aim for both a greater convergence Robustness is more important than ef®ciency. of underlying assumptions, and a more explicit and examined Robustness has been one of the great strengths of the Inter- discussion of those assumptions. net, integral to its design from the very beginning [Clark88]. In some sense, this note is in the tradition of Shenker et al.©s One of our overriding assumptions is that it is critical not to paper on ºPricing in Computer Networks: Reshaping the Re- subordinate robustness to the goal of more closely approxi- search Agendaº [Shenker96], which is a discussion about for- mating optimal ef®ciency.
    [Show full text]
  • Third Annual Workshop on Meteorological and Environmental Inputs to Aviation Systems
    National Aeronautics and Space Ad ministration NASA CP-2104 PROCEEDINGS: THIRD ANNUAL WORKSHOP ON METEOROLOGICAL AND ENVIRONMENTAL INPUTS TO AVIATION SYSTEMS APRIL 3-5,1979 UNIVERSITY OF TENNESSEE SPACE INSTITUTE EDITORS: DENNIS W. CAMP WALTER FROST FAA-RD-79-49 i APPROVAL PROCEEDINGS: THIRD ANNUAL WORKSHOP ON METEOROLOGICAL AND ENVIRONMENTAL INPUTS TO AVIATION SYSTEMS Edited by Dennis W. Camp and Walter Frost The i formation in this report has been reviewed for t chni a1 content. Review of any information concerning Department of Defense or nuclear energy activities or programs has been made by the MSFC Security Classification Officer. This report, in its entirety, has been deter- mined to be unclassified. ahcmCHARLES A. LUNDQUIST / Director, Space Sciences Laboratory TECHNICAL REPORT STANDARD TITLE PAGE 1 REPORT NO, 12 GOVERNNENT ACCESSION NO. 13 RECIPIENT’S CATALOG NO. Proceedings. Third Annual Workshop on Meteorological and Environmental Inputs to Aviation Systems 6 PERFORMING ORGANIZATION CODE The University of Tennessee Space Institute Tullahoma, Tennessee 37388 12 SPONSORING AGENCY NAME AND ADDRESS Conference Publicat ion istration, Washington, D C 20553 mospheric Administration, The proceedings of a workshop on meteorological and environmental inputs to aviation systems held at The University of Tennessee Space Institute, Tullahoma, Tennessee, Apr.11 3-5, 1979, are reported The workshop was jointly sponsored by NASA, NOAA, and FAA and brought together many disciplines of the aviation communities in round table discussions The
    [Show full text]
  • The Internet
    The Internet G iovanni N eglia U niversità di Palerm o 2006/2007 Slides by courtesy of prof. B ianchi G. Bianchi, G. Neglia Traditional approach to Internet Teaching 1. transmission technologies • physical carriers, modulation, etc 2. data link protocols • reliable transfer of bits from point to point 3. Packet switching • Historical perspective, then technologies, routing, protocols, finally IP 4. Packet forwarding • Glue IP routing with layer 2, ARP,... 5. Transport protocols, application protocols • In a rush!! (just a bit of TCP, HTTP, …) G. Bianchi, G. Neglia 1 A pproach adopted in this course (almost) Top-Down ° Applications are indeed important ° What you see is what you learn first ß Start focusing on internet application programming ° Notion of sockets (no Java programming this year) ° Transport layer as application developement platform ß Web as driving application ° Limited details on other apps G. Bianchi, G. Neglia Course objectives & limits ß O B J ECTIV ES : ° U nderstanding w hat type of netw ork the Internet really is. ° U nderstanding w hy protocols have been designed as they are ° achieving capability to respond to laym an (the m ost critical) questions ° know ing w hat to read, w hen tech problem s arise ß LIM ITS : ° Scope lim ited to “just” inter-netw orking; no netw orking (no m ention to w hat’s below the internet protocol – dealt w ith in past courses) ° Lim ited to basic classical Internet (no m ention to recent developem ents) G. Bianchi, G. Neglia 2 Teaching Material Book and notes ° Nicola Blefari Melazzi, dispense, versione 4.2 (in italian), 2003 • Available online • In progress (310 pages at the moment) ° James F.
    [Show full text]
  • Next‐Generation Firewall Performance Benchmarking Methodology Draft
    Next‐Generation Firewall Performance Benchmarking Methodology Draft draft‐balarajah‐bmwg‐ngfw‐performance‐02 IETF 101, London, March 20, 2018 Bala Balarajah / Carsten Rossenhövel Goals . Provide benchmarking terminology and methodology for next‐generation network security devices including . Next‐generation firewalls (NGFW) . Intrusion detection and prevention solutions (IDS/ IPS) . Unified threat management (UTM) . Web Application Firewalls (WAF) . Strongly improve the applicability, reproducibility and transparency of benchmarks . Align the test methodology with today's increasingly complex layer 7 application use cases Table of Contents (1) 1. Introduction 2. Requirements 3. Scope 4. Test Setup . Test Bed Configuration . DUT/SUT Configuration . Test Equipment Configuration 5. Test Bed Considerations 6. Reporting . Key Performance Indicators Table of Contents (2) 7. Benchmarking Tests 1. Throughput Performance With NetSecOPEN Traffic Mix 2. Concurrent TCP Connection Capacity With HTTP Traffic 3. TCP/HTTP Connections Per Second 4. HTTP Transactions Per Second 5. HTTP Throughput 6. HTTP Transaction Latency 7. Concurrent SSL/TLS Connection Capacity 8. SSL/TLS Handshake Rate 9. HTTPS Transactions Per Second 10. HTTPS Throughput Test Setup Aggregation Solution Aggregation Switch/Router Switch/Router (optional) Under Test (optional) Emulated Router(s) Emulated Router(s) (optional) (optional) Emulated Clients Emulated Servers Test Equipment Test Equipment Feature Profiles NGFW Initial NGFW Future NG‐IPS AD WAF BPS SSL Broker SSL Inspection x Intrusion (IPS/IPS) x Web Filtering X Antivirus x Anti Spyware x Anti Botnet x DLP x DDoS x Certificate Validation x Logging and Reporting x App Identification x Key Performance Indicator (KPI) Definitions . TCP Concurrent Connections . TCP Connection Setup Rate . Application Transaction Rate . TLS Handshake Rate .
    [Show full text]
  • Measurement and Modelling of Internet Traffic Over 2.5 and 3G Cellular Core Networks
    DISSERTATION Measurement and Modelling of Internet Traffic over 2.5 and 3G Cellular Core Networks ausgef¨uhrt zum Zwecke der Erlangung des akademischen Grades eines Doktors der technischen Wissenschaften eingereicht an der Technischen Universit¨at Wien Fakult¨at f¨ur Elektrotechnik und Informationstechnik von DI Philipp Svoboda Wanriglgasse 1/5 A-1160 Wien Osterreich¨ geboren am 25. M¨arz 1978 in Wien Matrikelnummer: 9825199 Wien, im November 2008 Begutachter: Univ. Prof. Dr. Markus Rupp Institut f¨ur Nachrichtentechnik und Hochfrequenztechnik Technische Universit¨at Wien Osterreich¨ Univ. Prof. Dr. Andreas Kassler Computer Science Department Karlstadt University Sweden Abstract HE task of modeling data traffic in networks is as old as the first commercial telephony systems. TIn the recent past in mobile telephone networks the focus has moved from voice to packet- switched services. The new cellular mobile networks of the third generation (UMTS) and the evolved second generation (GPRS) offer the subscriber the possibility of staying online everywhere and at any time. The design and dimensioning is well known for circuit switched voice systems, but not for mobile packet-switched systems. The terms user expectation, grade of service and so on need to be defined. To find these parameters it is important to have an accurate traffic model that delivers good traffic estimates. In this thesis we carried out measurements in a live 3G core network of an Austrian operator, in order to find appropriate models that can serve as a solid basis for traffic simulations. A requirement for this work was a measurement system, which is able to capture and decode network traffic on various interfaces of the mobile core network.
    [Show full text]
  • Multiplexing Traffic at the Entrance to Wide-Area Networks
    Multiplexing Traf®c at the Entrance to Wide-Area Networks RamoÂn CaÂceres Ph.D. Dissertation Report No. UCB/CSD 92/717 Computer Science Division University of California Berkeley CA 94720 December 1992 [email protected] -ii- Multiplexing Traf®c at the Entrance to Wide-Area Networks Copyright 1992 by RamoÂn CaÂceres - iii - A mis padres, Mon y Mirtha. -iv- Abstract Many application-level traf®c streams, or conversations, are multiplexed at the points where local-area networks meet the wide-area portion of an internetwork. Multiplexing policies and mechanisms acting at these points should provide good performance to each conversation, allocate network resources fairly among conversations, and make ef®cient use of network resources. In order to characterize wide-area network traf®c, we have analyzed traces from four Inter- net sites. We identify characteristics common to all conversations of each major type of traf®c, and ®nd that these characteristics are stable across time and geographic site. Our results contrad- ict many prevalent beliefs. For example, previous simulation models of wide-area traf®c have assumed bulk transfers ranging from 80 Kilobytes to 2 Megabytes of data. In contrast, we ®nd that up to 90% of all bulk transfers involve 10 Kilobytes or less. This and other ®ndings may affect results of previous studies and should be taken into account in future models of wide-area traf®c. We derive from our traces a new workload model for driving simulations of wide-area internetworks. It generates traf®c for individual conversations of each major type of traf®c.
    [Show full text]
  • UNITED STATES SECURITIES and EXCHANGE COMMISSION Washington, D.C
    UNITED STATES SECURITIES AND EXCHANGE COMMISSION Washington, D.C. 20549 - -------------------------------------------------------------------------------- FORM 10-K (Mark One) |X| ANNUAL REPORT PURSUANT TO SECTION 13 OR 15(d) OF THE SECURITIES EXCHANGE ACT OF 1934 [FEE REQUIRED] For the fiscal year ended December 31, 1996 OR | | TRANSITION REPORT PURSUANT TO SECTION 13 OR 15(d) OF THE EXCHANGE ACT OF 1934 [NO FEE REQUIRED] For the transition period from _______ to _______ Commission file number 0-19551 - -------------------------------------------------------------------------------- Atlantic Tele-Network, Inc. (Exact name of registrant as specified in its charter) Delaware Chase Financial Center (State or other jurisdiction of P.O. Box 1730 incorporation or organization) St. Croix, U.S. Virgin Islands (Address of principal executive offices) 00821 47-0728886 (Zip Code) I.R.S. Employer Identification No.) (809) 777-8000 (Registrant's telephone number, including area code) _______________________________________________ Securities registered pursuant to Section 12(b) of the Act: Name of each exchange on Title of each class which registered ---------------------------- ---------------------------- Common Stock, Par Value $.01 per Share American Stock Exchange Securities registered pursuant to Section 12(g) of the Act: Title of each class - ---------------------------- None _______________________________________ Indicate by check mark whether the registrant (1) has filed all reports required to be filed by Section 13 or 15(d) of the Securities
    [Show full text]
  • A Peer-To-Peer Internet for the Developing World SAIF, CHUDHARY, BUTT, BUTT, MURTAZA
    A Peer-to-Peer Internet for the Developing World SAIF, CHUDHARY, BUTT, BUTT, MURTAZA Research Article A Peer-to-Peer Internet for the Developing World Umar Saif Abstract [email protected] Users in the developing world are typically forced to access the Internet at a Associate Professor fraction of the speed achievable by a standard v.90 modem. In this article, we Department of Computer Science, LUMS present an architecture to enable ofºine access to the Internet at the maxi- Opposite Sector U, DHA mum possible speed achievable by a standard modem. Our proposed architec- Lahore 54600 ture provides a mechanism for multiplexing the scarce and expensive interna- Pakistan tional Internet bandwidth over higher bandwidth P2P (peer-to-peer) dialup ϩ92 04 205722670, Ext. 4421 connections within a developing country. Our system combines a number of architectural components, such as incentive-driven P2P data transfer, intelli- Ahsan Latif Chudhary gent connection interleaving, and content-prefetching. This article presents a Research Associate detailed design, implementation, and evaluation of our dialup P2P data trans- Department of Computer fer architecture inspired by BitTorrent. Science, LUMS Opposite Sector U, DHA Lahore 54600 Pakistan 1. Introduction The “digital divide” between the developed and developing world is un- Shakeel Butt derscored by a stark discrepancy in the Internet bandwidth available to Research Associate end-users. For instance, while a 2 Mbps ADSL link in the United States Department of Computer costs around US$40/month, a 2 Mbps broadband connection in Pakistan Science, LUMS costs close to US$400/month. Opposite Sector U, DHA The reasons for the order of magnitude difference in end-user band- Lahore 54600 Pakistan width in the developed and the developing world are threefold: • Expensive International Bandwidth: Developing countries often Nabeel Farooq Butt have to pay the full cost of a link to a hub in a developed country, Research Associate making the cost of broadband Internet connections inherently ex- Department of Computer pensive for ISPs.
    [Show full text]